Dayaks
[H]F Junkie
- Joined
- Feb 22, 2012
- Messages
- 9,776
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
NVIDIA GeForce GTX 1080 Founders Edition Review @ [H]
- Joined
- May 18, 1997
- Messages
- 55,651
Hard for us to make comparisons when we do not have the hardware in hand. Hell you can read in this thread alone, we made one opinion about Async Compute for AotS and someone gets their panties in a wad because we have no proof. If I had proof, it would not be an opinion, it would be a fact.
- Joined
- May 18, 1997
- Messages
- 55,651
Damn near silent, just like the 980Ti ref card.
Also it still seems the Async Compute is disabled within the DX12 render path for NVIDIA; Ielda tested using both the newer driver (meant to handle how to deal with the Async Compute issue and so bypassed by driver rather than application) and older one with the latest version of AoTS on Steam.
Result was more similar than different on their 980ti, meaning that the result does not really reflect what the 1080 will be able to do anyway (with regards if performs better in AoTS with async compute),
Wonder if NVIDIA will work with Oxide to have this re-enabled to test with Pascal.
Cheers
- Joined
- May 18, 1997
- Messages
- 55,651
Not sure who that is you are referring to, but reference cards have NEVER had better cooling solutions than high end AIB cards.....pretty much ever. I think you might be able to argue that ref cards from recent years have better circuitry, but maybe not capable of supplying more power cleanly than AIB cards. Ref cards are built like little tanks to run forever at stock speeds. Once you start taking those outside of that profile, all bets are off. Those cards are simply not designed for that.
I still do not get why they kept it at 64 ROPs like the 980 though, when it has more SM/Cuda Cores/Textures <----- Doh actually same number of cluster (GPC) as 980.
Surely it would had contributed to the performance, assuming bandwidth is fine (would be up to 1440p).
Cheers
Edit:
NVM, probably comes back to the cluster of 4 GPC, same as 980 and where the 980ti had 6.
Surely it would had contributed to the performance, assuming bandwidth is fine (would be up to 1440p).
Cheers
Edit:
NVM, probably comes back to the cluster of 4 GPC, same as 980 and where the 980ti had 6.
So that they could sell you a 1080Ti next year
I never really see my card so I don't care so much about looks (and the 1080 cooler looks a bit weird anyway). But yes if it performs dramatically better I'd be willing to pay a bit more. Maybe not $100 more, I don't know.
I really appreciate all the work nvidia put into these cards, but the pricing leaves me feeling meh for what you get. What was the 980, $550? And now they're asking $150 more for a Gx104 chip? They're asking more than the launch price of the 980 Ti, which was a full GM200?
Brent_Justice
Moderator
- Joined
- Apr 17, 2000
- Messages
- 17,755
It would raise them, there is a direct correlation between temperature and clock speed to some degree.
Brent_Justice
Moderator
- Joined
- Apr 17, 2000
- Messages
- 17,755
They are absolutely allowed to yes, I suspect we will see some dual 8pin designs.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Nvidia GeForce GTX 1080 im Test (Seite 11)
async compute tested
code being tested (n-body gravity)
nBodyD3D12/D3D12nBodyGravity.cpp at c657e94ff0b5ca3fd87a1c51900c8b295ac6ccc4 · GPUOpen-LibrariesAndSDKs/nBodyD3D12 · GitHub
Brent_Justice
Moderator
- Joined
- Apr 17, 2000
- Messages
- 17,755
It can actually perform as a passive heatsink, I noticed heat was spread across it quite evenly, and very hot to the touch.
- Joined
- May 18, 1997
- Messages
- 55,651
Please do NOT turn our review thread in another Async Compute argument.....do that in your own thread please, or one of the hundred others already posted.
Napoleon
[H]ard|Gawd
- Joined
- Jan 27, 2003
- Messages
- 1,073
Makes you wonder if AIBs might capitalize on that for some better passive cooling. I think these cards have huge upside
Correct question.
Check here for numbers of the 1080GTX 8GB Vs TitanX 12GB or R9 295x2 8GB : NVIDIA GeForce GTX 1080 8 GB Review
From reviews measuring fan speeds versus performance it's very clear the reference 1080 is constrained by temperature, not power. With much better AIB cooling that may change. Or it may not-- we don't know yet, maybe GP104 tops out around 2.1Ghz regardless.
- Joined
- May 18, 1997
- Messages
- 55,651
No offense, I do not know who that is, and quite frankly do not care. I would suggest you ask him questions about his opinions. If you have questions about our content on our forums, I am fully open to that discussion.
Digital Viper-X-
[H]F Junkie
- Joined
- Dec 9, 2000
- Messages
- 15,117
Interesting! Now I really look forward to the 1070, if it's 20% slower, that should put it @ 980TI levels
Last edited:
Brent_Justice
Moderator
- Joined
- Apr 17, 2000
- Messages
- 17,755
That 7GB of VRAM usage was in Tomb Raider at max in-game settings.
defaultluser
[H]F Junkie
- Joined
- Jan 14, 2006
- Messages
- 14,398
But I'm looking forward to the improved SLI scaling. Might have more of an effect on 4k than it does at 1440p, since ithe 1080 has faster card-to-card bandwidth AND more available memory bandwidth (thanks to compression).
I'm pretty sure the SLI bus is limiting SLI at 4k.
Very hot? That makes it even more difficult for those of us with mini-ITX cases.
Digital Viper-X-
[H]F Junkie
- Joined
- Dec 9, 2000
- Messages
- 15,117
Well.. regardless of the backplate, the heat will be there, just more concentrated on certain parts of the board.
Quoting myself here, but a german site reviewed the FE card with aftermarket cooling, the Accelero Xtreme IV. I can't speak German, but videocardz's conclusion from their video was that with great aftermarket cooling the 1080 is indeed constrained by power!
NVIDIA GeForce GTX 1080 tested with aftermarket cooler | VideoCardz.com
So when AIB cards with additional power connectors come along with cherry-picked GPUs and high-end air coolers, we can reasonably expect to get well past 2.1Ghz.
only when the 1080Ti comes out.. if anything playing DOOM makes me realise my 980Ti was an over purchase, I bought it at the same time as an XB270HU so g-synch also feels like an monumental over purchase too... but for the first time in my memory (going back to TNT2 with Nvidia) I have so much in reserve that I'm more than happy to wait a year.. roll on 1080Ti and 4k 144hz monitors
Dayaks
[H]F Junkie
- Joined
- Feb 22, 2012
- Messages
- 9,776
Seems a bit toasty for a cooler they're charging an additional $100 for.
I skipped Kepler and Maxwell so let's see if I can hold out for Volta.
It doesn't ramp up the fan until it gets near max temp. I read somewhere around here the fan only runs at 55%. Guru3d did their OC testing at 60% IIRC.
Heh, exactly what I said. Glad great minds think alike. I'm looking forward to that article.
Also, I was under the impression that a backplate adds to the rigidity of the card; less potential for bending. I personally prefer a card with a backplate over one without for aesthetics (hides the electronics). The passive cooling it provides as well is a bonus.
TPU so far has spent a fair bit of time looking at the behaviour and performance window of the Boost 3/temperature/clocking.
NVIDIA GeForce GTX 1080 8 GB Review
NVIDIA GeForce GTX 1080 8 GB Review
It is definitely more complex and a different approach than in the past, but also mentions current 1080FE restricted by the board power limit and temps.
Nothing new there because the 980ti reference was also if I remember correctly, just it now also behaves differently managing those variables.
I think Kyle/Brent/HardOCP is right to take the time on this one, especially as the right tools to do this for this new approach in Boost3 are still sort of in beta.
Cheers
Last edited:
Criticalhitkoala
[H]ard|Gawd
- Joined
- Apr 17, 2015
- Messages
- 1,986
I'm impressed. . When people talk about gen to gen performance on Intel chips only going up by 10-15% as low. I would imagine they would be happy as hell that there's a 30% increase. He'll comparing a 1080 to a 980 there is a massive gap. Good on nVidia. I'll be getting some 1080s. But waiting for Asus or zotacs implementation.
Yes, also good links. We know there's a spot for a second power connector on the 1080 reference cards. Since they're power-constrained, that's yet another reason not to buy the reference. Hold out for a few more weeks, friends! We can do it!
Axehandler
Gawd
- Joined
- Dec 19, 2007
- Messages
- 635
I'm still considering the 1070's SLI'd - curious how the new sli bridge will play into this.
Unfortunate that they are charging a premium for essentially nothing, that only punishes SFF builders. Great marketing spin though, that they charge $100 on something that performs worse than an EVGA AIO cooler which is also in the $100 ballpark.
Also they mentioned keeping reference cards in the supply chain for charging an extra $100. Except no one has trouble finding a reference 980 Ti a year after release. A complete manufactured "problem".
Also they mentioned keeping reference cards in the supply chain for charging an extra $100. Except no one has trouble finding a reference 980 Ti a year after release. A complete manufactured "problem".
- Joined
- May 18, 1997
- Messages
- 55,651
Some of the tools inside of Boost 3 that are going to need to be used are still in Alpha from the AIBs. NV has seemingly finished the structure, just waiting for AIBs to put their UI on those.
MavericK
Zero Cool
- Joined
- Sep 2, 2004
- Messages
- 31,910
I just went SLI with my 970s not long ago so this may seem odd, but I'd really like to get away from SLI. It seems like the support lately has been waning and even when it does work the gains are not that massive.
Maybe the new "2x bandwidth" bridge will help, but ultimately it comes down to driver/developer support which feels like it's been worse lately than in the past.
My real takeaway from that german review is that the reference cooler actually isn't a huge problem. You can sustain 2.1Ghz on the reference card with the fan at 55%. But that's your cap-- you can't get past 2.1Ghz because the reference card is power-constrained. The real point of AIB cards this generation might be that they come with another power connector.
According to Nvidia it seems like the new SLI bridge is only for the 1080.
"NVIDIA’s new SLI HB bridge doubles the available transfer bandwidth compared to the NVIDIA Maxwell™ architecture. Delivering silky-smooth gameplay, it’s the best way to experience surround gaming—and it’s only compatible with the GeForce GTX 1080."
That was taken from here: GeForce GTX 1080 Graphics Card
That's the one part that made zero sense to me. Even now you can hop on Newegg and purchase a Zotac 980ti with the reference Nvidia blower.
Nenu
[H]ardened
- Joined
- Apr 28, 2007
- Messages
- 20,315
If you overclock a lot and use the card a lot it can help the card last.
It gets extremely hot so it is needed.
I fitted an Accelero Xtreme IV to my 980ti, this has a much large backplate with a lot of fins.
This still gets extremely hot, even with 2 fans blowing on it.
If this wasnt dissipating that heat, something would be toasting a lot faster.
Without a backplate, if you moderately overclock and dont push the voltage, your card wont likely suffer during its life in your machine.
With the backplate you can get rid of more heat in the same timeframe.
This will help a higher overclock remain stable for longer gaming sessions.